Scene-by-scene checklist for a 45-second product explainer
A 45-second product explainer works when every scene has one job, one spoken line, and one must-show visual you approve before render.
Scene-by-scene checklist for a 45-second product explainer
A 45-second product explainer works when every scene has one job, one spoken line, and one must-show visual—and you can approve those stills before you render. Length is not the hard part; scene discipline is.
Forty-five seconds is roughly eight to ten short beats if you pack them, or five to six if you teach cleanly. Most SaaS explainers that feel rushed tried to do eight jobs. Use a scene checklist so the board stays teachable.
Time budget before you write frames
Decide the clock first. A practical split for 45 seconds:
| Block | Seconds | Job |
|---|---|---|
| Hook | 0–3 | Persona + friction + feature cue |
| Context | 3–10 | Why the old way fails |
| Mechanism | 10–28 | How it works (1–2 scenes) |
| Proof / edge | 28–38 | Believable detail (process, not fake %) |
| CTA | 38–45 | One next step |
If mechanism needs more than ~18 seconds, cut a feature—not the hook. Viewers who miss the opening never reach your clever middle.
Scene card template (copy for every beat)
Fill one card per scene. Do not generate motion until every card is complete.
- Scene # / name
- Seconds (start–end)
- Viewer question this scene answers
- VO line (≤18 words when possible)
- On-screen text (≤6 words)
- Must-show nouns
- Must-avoid (clutter, extra logos, second feature)
- Camera (hold / push / pan — pick one)
- Approval (yes / change / kill)
This card is the lightweight version of a full AI video storyboard checklist from concept to approved beats. Use whichever depth your reviewers need; the point is a shared definition of “done.”
Scene-by-scene checklist for a 45s SaaS explainer
Scene 1 — Hook (0–3s) - [ ] Role or persona is visible or spoken - [ ] Friction is concrete - [ ] Feature or outcome is named - [ ] Works muted - [ ] Safe crop for 9:16
Scene 2 — Context (3–10s) - [ ] Shows the failure mode of the old workflow - [ ] Does not introduce a second product claim - [ ] Visual metaphor matches the VO nouns - [ ] Ends with a bridge toward the mechanism
Scene 3 — Mechanism A (10–20s) - [ ] One step only - [ ] UI or metaphor is readable at phone size - [ ] No tour of the whole app - [ ] VO explains what happens, not every button
Scene 4 — Mechanism B (optional, 20–28s) - [ ] Only exists if Scene 3 cannot hold the idea alone - [ ] Continues the same metaphor family - [ ] Still one idea
Scene 5 — Proof / edge (28–38s) - [ ] Uses process proof (labels, steps, constraints)—not invented benchmarks - [ ] Honest about who it is for - [ ] Does not reopen the hook’s pain as a new rant
Scene 6 — CTA (38–45s) - [ ] One action - [ ] Matches the promise in Scene 1 - [ ] On-screen CTA ≤4 words when possible - [ ] No “link in bio + demo + newsletter” stack
Mapping product complexity into six scenes
If the feature has three sub-capabilities, do not show all three in one Reel. Pick the capability that answers the hook’s friction. Park the others for a follow-up clip. Educators facing the same split problem when turning lessons into shorts use a similar cut discipline in from course outline to short tutorial clips.
A useful filter: Would deleting this scene break the promise line? If no, delete it from this cut.
Silent comic-strip QA (before any render)
Print or scroll the stills with audio off:
- Can you retell the product promise from frames alone?
- Does each still answer its assigned viewer question?
- Are any two scenes doing the same job?
- Is the CTA readable in the last still?
Failures here are cheaper than regenerations after motion. The pattern matches storyboard-first vs generate-then-fix: plan until approval is boring, then render once.
Mini filled example: “shared call annotations”
Promise: After this Reel, a CSM can pin a moment and share it without rewatching the call.
| Scene | VO | Must-show |
|---|---|---|
| 1 Hook | CSMs: stop rewatching the whole call. | Scrubbing timeline fatigue |
| 2 Context | Notes in a doc lose the moment. | Doc + missed timestamp |
| 3 Mechanism | Pin the second that matters. | Pin on waveform |
| 4 Mechanism | Share the clip with context. | Shared card with note |
| 5 Proof | Same labels your team already uses. | Familiar tag chips |
| 6 CTA | Try a pin on your next call. | Single CTA card |
Six cards. Forty-five seconds. No invented “3× productivity” claims.
Common checklist skips that force regenerations
- Hook still approved, but VO rewritten after render
- Mechanism scene with three UI panels at once
- Proof scene that invents a metric
- CTA that introduces a new feature
- Aspect ratio decided after storyboard lock
Those skips show up again and again in common storyboard mistakes that waste AI video credits. Put the checklist upstream of the credit spend.
After approval: render and compose
Render only approved scenes. Compose captions from the locked VO. Generate a cover still that survives platform crops (hook thumbnails and cover stills). Then measure whether viewers actually learned the concept—views alone do not prove teaching.
Soft tool note
A shared board where marketers can approve scene cards before motion keeps 45-second explainers honest. Scro is one place to keep stills, VO, and review in one loop; the checklist itself is stack-agnostic.